Deno Team Joins Cloudflare: What It Means for Edge Runtime Lock-In

What Cloudflare announced

Cloudflare announced on October 9, 2026, that the Deno team is joining the company. The stated goal is to merge code and ideas from Deno’s celld project into Cloudflare’s open-source workerd runtime, making self-hosting Workers and Durable Objects a first-class supported way to build and run applications.

This is a vendor-reported plan, not a completed product release. Ryan Dahl and Bert Belder will lead the new effort, and Cloudflare says more announcements are expected in the coming months. The evidence does not establish when distributed self-hosting will reach production readiness, which environments it will support first, or whether existing Workers applications will move without changes.

Why Durable Objects matter to portability

Cloudflare describes Durable Objects as individually addressable servers with their own relational SQLite databases. Their JavaScript execution is single-threaded, and they can handle WebSockets while accessing local SQLite synchronously. In Cloudflare’s example, assigning one Durable Object to each chat channel can shard data and WebSocket connections across machines.

That model changes what an application expects from its runtime. A team is not only moving stateless request handlers between hosts. It may also need to reproduce object placement, routing, durable storage, and coordination across machines. A compatible runtime alone does not recreate the surrounding operational system.

Cloudflare says workerd is open source and implements Durable Objects, but its open-source Durable Objects support has been limited to a single instance. The company describes that implementation as good enough for local testing but unable to scale across machines. Cloudflare’s production routing system was designed for a network spanning hundreds of locations and depends on infrastructure operated by reliability engineers. The company says that system is not the implementation a self-hosting operator would want.

Deno’s celld project was aimed at that gap. Cloudflare describes celld as one Rust binary whose only external service dependency is object storage. Operators manage multiple celld instances and one object storage bucket, while the runtime provides compute, relational databases, queues, and real-time communication primitives for distributed applications.

What the deal does and does not solve

The proposed merge could make the Workers programming model easier to reproduce outside Cloudflare’s managed network. Developers who value Durable Objects would have a clearer path to running compatible primitives on infrastructure they operate themselves. That could reduce one important form of migration friction: rebuilding the runtime model when an application depends on stateful, distributed objects.

Workers would still not become identical to traditional cloud platforms. Cloudflare argues that Workers is different because its programming model, bindings, and Durable Objects were designed together. The company also says that being open source provides an escape hatch, while acknowledging that workerd has not yet had the adjacent services and tooling needed for broad production use.

Teams should separate three questions. Can their application run on the open-source runtime? Can the self-hosted system provide distributed Durable Objects with the required routing and storage behavior? Can their operations team support the resulting cluster, backups, network paths, and object storage dependency? The announcement addresses the second question as an active engineering target, not as a finished guarantee.

What SiteList’s index can and cannot show

SiteList’s Infrastructure & hosting category contains 23 published listings. Across the 23 scored listings, the median SiteList Score is 76.8, with scores ranging from 31.7 to 84.

A SiteList Score is a crawler-based review of a product’s public website and documentation across 13 dimensions. It does not measure runtime performance, uptime, security, adoption, or customer satisfaction. These figures describe SiteList’s reviewed listings, not the infrastructure market as a whole.

The category evidence matters here because portability decisions depend on public documentation as well as runtime design. A vendor can publish an open-source runtime while leaving operators to infer how placement, routing, storage, upgrades, and failure recovery work. The category range from 31.7 to 84 shows that the public website and documentation evidence available across the reviewed infrastructure listings varies widely. It does not prove that any particular product is easier or harder to operate.

Engineering teams comparing Cloudflare’s plan with another infrastructure option should inspect the specific migration boundary: supported APIs, state placement, storage dependencies, local testing, cluster operations, and backup procedures. The category median is context for that review, not a substitute for it.

What teams should watch next

Developers can start self-hosting celld or workerd today, according to Cloudflare, but the announced initiative is intended to make workerd self-hosting a first-class supported path. Teams with an immediate requirement should treat current implementations as separate technical options and verify their own workloads before committing to a migration plan.

The most important future evidence will be concrete documentation. Teams should look for a supported deployment model, a clear explanation of distributed Durable Object routing, storage and recovery guidance, compatibility details, and dated examples showing how an application moves between Cloudflare’s managed environment and self-hosted infrastructure.

Cloudflare says the announcement makes portability more credible as a direction. It does not yet make portability automatic. For engineering leaders concerned about lock-in, the practical test will be whether the merged project exposes enough of the operational model to run, debug, upgrade, and recover the same application outside Cloudflare.